Early Prediction of Student Performance in Online Programming Courses.

Early prediction of student grades is important for both teachers and students. It can help teachers take timely remedial actions to avoid drop-out and poor learning outcomes, and also help students improve their engagement, motivation and achievement of desired results. In this study, we present a machine learning approach to predict the final student grade from information available in the middle of the course, in the context of introductory programming courses for primary and high school students. We define and extract suitable features from the raw data and use a decision tree classifier to produce a compact set of rules, which are both accurate and interpretable by teachers and students. The decision tree rules provide insights about the important factors for success and highlight key programming tasks that predict the final student performance.